Author: Vincent Diepeveen
Date: 18:42:23 09/11/01
Go up one level in this thread
On September 11, 2001 at 03:05:20, Bruce Moreland wrote: When using singular extensions, extending all checks, most likely also extending passers (who knows?) and searching fullwidth, researching is endless. So after it failed low, god knows how long it would have taken fullwidth to get a new PV to just fail high for a better move. Each move they research somehow for singularity and that new window they need for it is completely different. That's worse than a standard alfabeta search, that's more closer to minimax depending upon what reduction was used for Singular Extensions. >On September 10, 2001 at 22:30:30, K. Burcham wrote: > >> [D] 4r3/8/2p2PPk/1p1r4/pP2p1R1/P1B5/2P2K2/8 b - - >> >> >>after the above position deep blue blundered with 44. ...Rd1. >>there was an immediate jump of 4 points with this move. >>not sure if this was a move mistake, slip, not sure. >>but with 44. ...g8 or 44. ...f5+ black is still 2.5 down >> with a losing position. i played this out with tiger 14 >> and ended with this position. mate in 9 with the f and g pawns. > >It is "blunders" that might provide us with insight. > >If a computer gives a knight in a manner that can be seen in a 2-ply search, it >should indicate that moves that appeared better also lost a knight, or perhaps >more. > >So it's possible that these other moves (Rg8, Rf5+) are refuted pretty deeply. >I have a hard time believing that DB just threw away material for no reason. > >bruce
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.